[Improve] Make completion reports easier to trust

[roomote-roomote]

​Opened on behalf of @mrubens. Follow up by mentioning @roomote-roomote, in the web UI, or in Slack.

What changed

Standard implementation tasks now ask completion reports to make four trust boundaries clear in natural prose: what completed, the evidence supporting it, the scope actually covered, and any material item not verified or still pending. A focused prompt-contract test keeps those distinctions explicit without introducing required headings or boilerplate.

Why this change was made

A fresh external signal from coding-agent builders emphasized that completion reports are easier to trust when claims separate outcomes, proof, scope, and verification gaps. Roomote already prefers concise conversational closeouts, so the smallest useful change tightens that existing contract instead of adding a report template or new runtime machinery.

Impact

• Scope: This changes only the Standard implement-changes final-report guidance and its regression coverage. Fast mode is intentionally excluded because open PR [Improve] Make Fast mode evidence-driven #1477 already covers evidence-driven Fast behavior.
• Evidence: The focused Vitest suite, @roomote/cloud-agents typecheck, targeted formatting check, and repository pre-push lint/type/knip gates passed.
• Not verified: No live-model evaluation was run, so adherence by every model/provider remains probabilistic. Browser proof is not applicable because the change has no rendered UI behavior.

[roomote-community]

No code issues found. See task

_{Reviewed 1005e06}